Xiaoyang Wu

That case is a bit out of my expectations. As four GPUs are predicting concurrently, it should not merely 17 `npy`. Yet this case means the other 3 GPUs are...

@aldinorizaldy Could you share your script for testing? It was directly started with `python` or our test script?

Sadly, In my local machine, the issue seems that did not occur. As shown in the below image, it is still the second iteration of testing, we already have multiple...

Hi, I noticed that "Totally 1 x 1 samples in Area_1 set.", Is that an irregular?

So, what's the batch size? If the batch size is even larger than the number of samples, an error may occur.

Here (https://github.com/Pointcept/Pointcept/blob/main/configs/scannet/semseg-ppt-v1m1-0-sc-st-spunet.py#L41) we maintain a union set of all class_name from each dataset. Then the "valid_index" for one of these datasets indicates the location index of the given class name...

> Well, in my case, I only have 2 `npy` in the second iteration. That's odd, but now it's clear where the problem is. Anyway, thanks for checking it. I'll...

@aldinorizaldy [[email protected]](mailto:%[email protected])

So, just debugging, how about increasing the number of samples?

> Hello @Gofinge, thank you for your answer. > > At the first glance we didn't catch that much of a difference so it seems we can use it without...